1. Technical Field
The present disclosure relates to image analysis devices.
2. Description of Related Art
Certain organizations need to monitor and capture photos/images from a particular area, such as a crossroads, via a camera. Each captured image may be analyzed to extract a background model and the image in the foreground. The background model includes those objects that would remain unaltered in a number of images, such as buildings and trees. The foreground image includes those objects that would change from image to image, such as people and cars. Comparison between the foreground images in a number of captured images is used to monitor the conditions and circumstance of the particular area being monitored, such as the quantity and the speed of the cars, and the quantities of pedestrians at different times.
The common method for analyzing the captured image to extract the background model and the foregrounds is to analyze one particular captured image to predetermine one scene as a reference background model for the particular area, and then compare the later captured images against the reference background model to determine the parts which are the same as the reference background model, and then filter out the parts which are the same, to obtain the foreground image from each captured image. However, a single reference background model does not take account of changes in the environment, such as changes in the levels of natural or other light illuminating the monitored area. Then, the background characteristics of the later captured images may be substantially different from those of the single background reference model, which results in the incorrect filtering or extraction of foreground images from the later captured images, or at least the need to adjust each later captured image before a worthwhile foreground image can be obtained from the later captured images.
Therefore, it is desirable to provide an image analysis device and method to overcome the described limitations.